"The newest version of ChatGPT passed the US medical licensing exam with flying colors — and diagnosed a 1 in 100,000 condition in seconds"

[1NJParent]

I’m no expert in medicine, but I suspect that a rare disease would be easy to be misdiagnosed. The Harvard doctor chose this particular rare disease presumably because it was hard to diagnose. The article only mentioned this one real-life case, but I’d assume the doctor had run many other cases through the AI system to reach his conclusion that “GPT-4 has better clinical judgment than ‘many doctors’”.